Palette reordering for quantised image compression through generic optimisation

Colour quantised images consist of a colour palette which contains all colours in the image and an index image which still constits of indices into the palette to make up the image. The index image still contains a large amount of redundant information and can therefore be compressed using lossless image compression algorithms. The achieved compression ratio however depends on the ordering of colour in the palette. In this paper we show that a generic optimisation technique such as simulated annealing can be used reorder the palette entries for subsequent JPEG2000 lossless compression of the index information, and demonstrate that our approach allows for improved compression compared to previous reordering techniques.